- Shivraj Asthana
AI-02: Big Data & Artful Intelligence
Catch me if you can
Yes, do cast a wide net.
Data is everywhere ... all around you - can you catch 'em - but do focus on quality rather than quantity.
You have heard this before: We begin with the three V’s: Volume, Velocity & Variety
(that's the do-re-me of Big Data)
Large Volumes of Structured data …
It’s the usual & the-not-very-exciting-stuff that lives in your internal databases; also in the public clouds. Always useful to have some outside data; gives you the ability to see how you compare with the outside trends.
To this we added the vast Variety of exciting Un-structured data.
Go ahead and happily include social perceptions.
Again it is critical to identify the ones that really matter to your business.
I can almost imagine some interesting ones.
Seriously, was this tweeted?
Really, what was posted regarding this?
How does Gen Y perceive this?
That status is so not cool!
What was pinned?
Yes, that too – the video that went viral?
Of course, with a high Velocity, you are almost capturing real-time data as well.
The in-memory columnar format – even better! Who wants a data-warehouse when you can have it alongside your transaction data.
Part or Whole …
Do you want to start with a Holistic set of data to get the big picture for your business?
Do you want to start with a small set of Deeper, Segmented data (think: specialized industries here)?
Analyze this …
What story does this data tell you?
Connect the dots and analyze from the business perspective.
Apply algorithms to derive intelligence and insights.
Missing Pieces …
Remember though to have a good data representation. Could there be any missing data that will affect the inference? What about the boundaries? Can't provide a skewed perspective, now can we?
Lay it out …
Time to provide some visually compelling and easy to understand insights for decision making.
Fast and Furious …
To this ever-changing data apply an evolving logic - logic that progressively learns from its experiences and fine-tunes in an ongoing manner.
You got it ... the logic applied should simply not be hard-coded ... it's smart enough to learn from itself... and does not require any manual logic upgrades. How do we manage that? Bingo!
Usher in - Innovative Cognitive Computing!
(… so la ti do …)